Creating Automatic Semantic Device Descriptions for Brownfield Industrial Robots
Abstract
Semantic device descriptions (SDD) are formal models that map signals from field devices such as sensors, actuators, and programmable logic controllers to a meaningful system model. Therefore, a SDD provides explicit identification of signals and their meaning, e.g., motor temperatures or robot axis positions. With the increasing use of signals for higher-level automation, SDDs become more important. While SDDs are usually specified in advance for greenfield systems, they often have to be defined manually for brownfield systems. This paper proposes a novel approach to automatically generate SDD from signal channels of device controls. Our approach uses an artificial neural network to assign sensor signals from a robot control (e.g., joint angle) to its provenance (e.g., axis). By letting the robot follow a distinct motion path, the neural network can identify unique signal patterns and assign the correct signal to its component of provenance. A subsequent module transforms the model’s predictions into an SDD for the device. We evaluate our approach on two industrial robots.
Citation: J. Ehrhardt, A. Nordhausen, Alexander Guhl, Marcel Lewke, Constantin Hildebrandt, Oliver Niggemann, “Creating Automatic Semantic Device Descriptions for Brownfield Industrial Robots,” SAMPE Europe Conference.